A Novel Application of Support Vector Machines to Detect Targets

Abstract

In this paper we develop a novel pulse radar detection scheme using Support Vector Machines (SVMs). SVMs are a powerful tool for pattern classification. Exploiting this, we design a SVM for pulse radar detection of targets embedded in noise. An adaptive pre-processing stage is included when the echoes are very weak. It is observed that the signal-to-sidelobe ratio obtained is much higher compared to neural network based pulse radar detection schemes. We further examine the noise tolerance, range resolution ability and Doppler tolerance of the new algorithm for pulse compression, and compare it with the ones available in literature.